Correlated timing noise and high precision pulsar timing: Measuring frequency second derivatives as an example

B. W. Stappers; C. Bassa; M. J. Keith; X. J. Liu

arxiv: 1907.03183 · v1 · pith:3PGSHWOZnew · submitted 2019-07-06 · 🌌 astro-ph.HE

Correlated timing noise and high precision pulsar timing: Measuring frequency second derivatives as an example

X. J. Liu , M. J. Keith , C. Bassa , B. W. Stappers This is my paper

Pith reviewed 2026-05-25 01:21 UTC · model grok-4.3

classification 🌌 astro-ph.HE

keywords pulsar timingred timing noisedispersion measure variationssecond frequency derivativeBayesian analysismillisecond pulsarstiming arraysgravitational waves

0 comments

The pith

Pulsar timing recovers the second spin frequency derivative despite red noise using Bayesian power-law models.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that the second derivative of a pulsar's spin frequency remains measurable even when red timing noise and dispersion measure variations are present in the data. Bayesian fitting jointly models these noise processes as power laws alongside the timing parameters, allowing recovery of injected values in simulations even when the noise models do not exactly match between injection and recovery. This matters because accurate measurement of the second derivative can reveal effects such as a pulsar's radial velocity, and it affects how timing arrays search for gravitational waves. The work also derives how the uncertainty on this derivative scales with the length of the observing baseline, depending on the spectral index of the noise.

Core claim

We show that the second frequency derivative can be recovered from times-of-arrival data that include red timing noise and dispersion measure variations when these are modeled as power laws within a Bayesian framework. Recovery works even if the noise model used for injection differs from the one used for recovery. The measurement uncertainty decreases with timing baseline T as T to the power of -7/2 plus alpha over 2 for shallow indices between 0 and 4, and as T to the power of -1/2 for steep indices above 8. When applied to European Pulsar Timing Array and Parkes Pulsar Timing Array data for 49 millisecond pulsars, statistically significant values appear for PSR B1937+21 and two others.

What carries the argument

Joint Bayesian fitting of timing parameters together with power-law models for red timing noise and dispersion measure variations

If this is right

The uncertainty on the second frequency derivative improves with longer timing baselines according to the power-law scaling that depends on the noise spectral index.
Statistically significant second frequency derivatives should be included in the ephemerides for PSRs J0621+1002, J1022+1001, and B1937+21.
Extended ELL1 models must be used for binary pulsars with small orbital eccentricities to avoid computational problems.
The presence and measurability of second frequency derivatives affect the sensitivity of pulsar timing arrays to gravitational waves.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the second frequency derivative traces radial velocity, the method could enable new measurements of pulsar distances or velocities once longer baselines are available.
The scaling relation for uncertainty could guide the design of future observing campaigns by showing how much additional time is needed to reach a target precision for a given noise spectrum.
The same joint-modeling approach might be tested on searches for higher-order frequency derivatives or other deterministic signals in existing timing array datasets.

Load-bearing premise

The red timing noise and dispersion measure variations are stationary power-law processes whose indices and amplitudes can be jointly fitted with the timing parameters without introducing bias in the recovered second frequency derivative.

What would settle it

A simulation or real dataset with a known injected or independently measured non-zero second frequency derivative where the Bayesian recovery yields a value inconsistent with the input within the reported uncertainties would falsify the recovery claim.

read the original abstract

We investigate the impact of noise processes on high-precision pulsar timing. Our analysis focuses on the measurability of the second spin frequency derivative $\ddot{\nu}$. This $\ddot{\nu}$ can be induced by several factors including the radial velocity of a pulsar. We use Bayesian methods to model the pulsar times-of-arrival in the presence of red timing noise and dispersion measure variations, modelling the noise processes as power laws. Using simulated times-of-arrival that both include red noise, dispersion measure variations and non-zero $\ddot{\nu}$ values, we find that we are able to recover the injected $\ddot{\nu}$, even when the noise model used to inject and recover the input parameters are different. Using simulations, we show that the measurement uncertainty on $\ddot{\nu}$ decreases with the timing baseline $T$ as $T^\gamma$, where $\gamma=-7/2+\alpha/2$ for power law noise models with shallow power law indices $\alpha$ ($0<\alpha<4$). For steep power law indices ($\alpha>8$), the measurement uncertainty reduces with $T^{-1/2}$. We applied this method to times-of-arrival from the European Pulsar Timing Array and the Parkes Pulsar Timing Array and determined $\ddot{\nu}$ probability density functions for 49 millisecond pulsars. We find a statistically significant $\ddot{\nu}$ value for PSR\,B1937+21 and consider possible options for its origin. Significant (95 per cent C.L.) values for $\ddot{\nu}$ are also measured for PSRs\,J0621+1002 and J1022+1001, thus future studies should consider including it in their ephemerides. For binary pulsars with small orbital eccentricities, like PSR\,J1909$-$3744, extended ELL1 models should be used to overcome computational issues. The impacts of our results on the detection of gravitational waves are also discussed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The scaling relation for ν̈ uncertainty under power-law noise is the clearest new piece here, backed by recovery tests, though the real-data claims rest on an unverified stationarity assumption.

read the letter

The paper's main contribution is the explicit scaling γ = −7/2 + α/2 for the uncertainty on the second spin-frequency derivative under shallow power-law red noise, plus the demonstration that injected ν̈ can be recovered even when the analysis model differs from the injection model. They also report posterior distributions for 49 EPTA and PPTA millisecond pulsars and flag statistically significant ν̈ in three cases. The simulation recovery tests across mismatched power-law parameters are the strongest part of the work; they show the method is not obviously fragile to moderate model error within the assumed family. The scaling itself follows from the covariance matrix properties rather than a fit to the target values, so it is reproducible in principle. The real-data application is straightforward and directly relevant to PTA noise budgets for nanohertz gravitational-wave searches. The soft spot is exactly the one flagged in the stress-test note. The simulations only explore power-law mismatches; they do not test spectral breaks, non-stationarity, or excess low-frequency power that could be absorbed into the cubic term. The paper fits the same power-law model to the real residuals without reporting per-pulsar checks such as posterior-predictive spectra or free-spectrum comparisons. That leaves open the possibility that the three reported significant ν̈ values are partly noise artifacts. The assumption of stationary power-law processes is standard in the field, but it is load-bearing here and not independently verified for the pulsars that drive the conclusions. This is a paper for PTA analysts and timing-array noise modelers. A reader already working on red-noise mitigation or second-derivative measurements will find the scaling and the specific pulsar results worth checking. It is coherent on its own terms and shows honest engagement with the existing Bayesian timing framework, so it deserves a serious referee even if revisions are needed on the validation side.

Referee Report

3 major / 2 minor

Summary. The paper claims that Bayesian modeling of pulsar timing residuals with power-law red noise and DM variations allows recovery of injected second frequency derivatives ν̈ even under mismatched noise models between simulation and recovery. It derives an analytic scaling for the uncertainty on ν̈ with timing baseline T of the form T^γ with γ = −7/2 + α/2 (0 < α < 4) or T^−1/2 (α > 8), demonstrates this via simulations, and reports statistically significant ν̈ detections for three pulsars (B1937+21, J0621+1002, J1022+1001) in EPTA/PPTA data.

Significance. If the power-law stationarity assumption holds, the simulation-based recovery results and the explicit scaling relation constitute a useful contribution to high-precision timing analyses, particularly for assessing biases in gravitational-wave searches. The explicit demonstration that ν̈ can be recovered under model mismatch is a concrete strength; the real-data posteriors for 49 pulsars provide a practical data product.

major comments (3)

[§3] §3 (simulations): recovery of injected ν̈ and the scaling γ = −7/2 + α/2 are shown only for stationary power-law mismatches; the manuscript does not test non-power-law spectra, spectral breaks, or non-stationarity, which directly affects whether the scaling and real-data ν̈ values remain unbiased.
[§4] §4 (EPTA/PPTA application): for the three pulsars reported with significant ν̈ (B1937+21, J0621+1002, J1022+1001), no per-pulsar validation is presented that the fitted power-law model adequately describes the residuals (e.g., no posterior-predictive spectrum or free-spectrum comparison); excess low-frequency power could be absorbed into the cubic term and bias the reported ν̈.
[Table 1] Table 1 or equivalent real-data table: the fitted α values for the pulsars with significant ν̈ are not tabulated or discussed relative to the regime boundaries (α < 4 vs. α > 8) used to quote the scaling; this leaves the applicability of the derived uncertainty scaling to the actual measurements unclear.

minor comments (2)

[Methods] Notation for the power-law index α is introduced without an explicit equation reference in the methods; adding Eq. (X) for the covariance kernel would improve clarity.
[Figure 3] Figure 3 (scaling plot): the transition at α ≈ 4–8 is shown but the error bars on the simulated points are not discussed; confirming they are consistent with the analytic γ would strengthen the figure.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive report. We respond point-by-point to the major comments below.

read point-by-point responses

Referee: [§3] §3 (simulations): recovery of injected ν̈ and the scaling γ = −7/2 + α/2 are shown only for stationary power-law mismatches; the manuscript does not test non-power-law spectra, spectral breaks, or non-stationarity, which directly affects whether the scaling and real-data ν̈ values remain unbiased.

Authors: We agree that the simulations and analytic scaling are derived exclusively under the stationary power-law assumption for the red noise. This matches the noise model used throughout the manuscript and is the standard assumption in many PTA analyses. The explicit demonstration of recovery under mismatched power-law indices is the intended scope; extending the tests to spectral breaks or non-stationarity would require a separate study with different noise parametrizations. We will add an explicit statement of this limitation in the discussion section. revision: partial
Referee: [§4] §4 (EPTA/PPTA application): for the three pulsars reported with significant ν̈ (B1937+21, J0621+1002, J1022+1001), no per-pulsar validation is presented that the fitted power-law model adequately describes the residuals (e.g., no posterior-predictive spectrum or free-spectrum comparison); excess low-frequency power could be absorbed into the cubic term and bias the reported ν̈.

Authors: The manuscript applies the standard power-law red-noise model used in the EPTA and PPTA pipelines. While per-pulsar posterior-predictive checks are not shown, the Bayesian evidence and posterior widths already incorporate the model assumptions. For the three pulsars with significant ν̈ we will add a short discussion noting that the recovered α values are consistent with the data and that any unmodelled low-frequency power would be a general limitation of the power-law assumption rather than specific to these sources. revision: partial
Referee: [Table 1] Table 1 or equivalent real-data table: the fitted α values for the pulsars with significant ν̈ are not tabulated or discussed relative to the regime boundaries (α < 4 vs. α > 8) used to quote the scaling; this leaves the applicability of the derived uncertainty scaling to the actual measurements unclear.

Authors: The posterior distributions for α are obtained for all 49 pulsars and are available from the analysis; however, they are not explicitly tabulated for the three significant detections nor compared to the regime boundaries. We will revise the manuscript to include the median α (with uncertainties) for B1937+21, J0621+1002 and J1022+1001 and state which scaling regime applies to each. revision: yes

Circularity Check

0 steps flagged

No significant circularity; scaling derived analytically from covariance matrix

full rationale

The paper derives the uncertainty scaling γ = −7/2 + α/2 directly from the analytic properties of the stationary power-law covariance matrix in the Bayesian timing likelihood (not by fitting to recovered ν̈ values). Simulations test recovery of injected ν̈ under model mismatch but do not calibrate the scaling formula itself. Real-data ν̈ PDFs for the 49 pulsars are obtained by direct application of the same likelihood; no self-citation chain, self-definition, or renaming of a fitted result is load-bearing for the central claims. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The analysis rests on the assumption that timing noise is a stationary Gaussian process with a power-law spectrum; this is a domain-standard modeling choice rather than a new postulate. No new entities are introduced. Two free parameters per noise process (amplitude and index) are fitted jointly with the timing parameters.

free parameters (2)

red-noise amplitude and spectral index
Fitted per pulsar to the timing residuals; the recovered ν̈ depends on these fitted values.
DM-variation amplitude and spectral index
Fitted jointly with red noise; affects the covariance matrix used for ν̈ inference.

axioms (1)

domain assumption Timing residuals are a linear combination of deterministic timing model plus stationary Gaussian noise with power-law spectrum.
Invoked throughout the Bayesian modeling section and simulation setup.

pith-pipeline@v0.9.0 · 5908 in / 1499 out tokens · 17333 ms · 2026-05-25T01:21:26.737483+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Constraints on the Primordial Black Hole Abundance using Pulsar Parameter Drifts
astro-ph.CO 2026-04 unverdicted novelty 8.0

The first search for scalar-induced gravitational waves via pulsar parameter drifts yields f_PBH < 10^{-10} (95% CL) for PBH masses 0.3 to 4e4 solar masses, strongly disfavoring a primordial black hole origin for LVK ...